By an image recognition technique, a computer is able to recognize various subjects such as a person, a face, a product, an animal, a vehicle, an obstacle, a character, and a two-dimensional code from an image. Various improvements are made in order to enhance recognition precision of recognition processing as described above. For instance, PTL 1 proposes a method, in which an amount of exposure of a plurality of cameras is adjusted in such a manner that the number of disparity values calculated regarding a stereoscopic object to be recognized increases in order to enhance recognition precision of the stereoscopic object. Further, PTL 2 describes an improvement on a feature pattern stored in a dictionary in order to enhance precision of individual recognition of a face. Specifically, in a configuration described in PTL 2, a feature pattern obtained by combining a feature of a target portion on a face area of a subject, and a feature of a portion other than the target portion is stored in a dictionary as a feature pattern for identifying the subject.